Image Quality Comparison: Radeon Gallium3D vs. Catalyst

Coming up in the next few days will be benchmarks of Mesa 8.0 with Morphological Anti-Aliasing (a.k.a. MLAA) plus some other imaging-oriented work/announcements to come in the near future. With that said, this weekend prior to leaving for Munich I ran some tests of the Radeon Gallium3D and Catalyst drivers when comparing the image quality.

This article is just a first-look from Mesa 8.0 and the proprietary Catalyst driver with two supported graphics cards. As a bonus is also a Radeon X1800XL on the R300g driver to show how the image quality compares to the R600g driver.

The screenshots in this article are of Nexuiz since there is the Nexuiz Image Quality test profile on OpenBenchmarking.org for use by the Phoronix Test Suite. While it's the raw performance, power consumption, CPU usage, or other metrics that I commonly look at on Phoronix.com, the Phoronix Test Suite has full support as well for doing image quality comparisons. It can also look for pixel-by-pixel differences, among other features, with the Phoronix Test Suite. I implemented all of that back in 2009 as talked about in this article, but it's just not used so often in the public spotlight. For example, anyone can run their own image quality comparison by running phoronix-test-suite benchmark nexuiz-iqc.

Anyhow, first up are some Nexuiz screenshots with a Radeon X1800XL on Mesa 8.0 R300g, Radeon HD 5770 on Mesa 8.0 R600g, and the Radeon HD 5770 on Catalyst fglrx 8.91.4 as found in Ubuntu 12.04 LTS. There's some subtle differences if looking closely.

Now to end off this very brief preview, here's some screenshots of the Radeon HD 6570 graphics card on Mesa 8.0 R600g and Catalyst fglrx 8.91.4 from the same software stack as used in the other tests. Additionally, since this is just a two-way comparison, the third render was generated by the Phoronix Test Suite analyze-image-delta option. This analytical option compares the two images on a pixel-by-pixel basis and highlights differences on the third composite.